[
https://issues.apache.org/jira/browse/SOLR-11597?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16308148#comment-16308148
]
Michael A. Alcorn commented on SOLR-11597:
------------------------------------------
The additional tests look good to me, [~cpoerschke].
Thanks for the feedback, [~yuyano].
bq. The name of "nonlinearity" is a bit curious for me because we call these
functions as "activation function". How about using "activation" instead of
''nonlinearity" to represent it?
"Nonlinearity" and "activation function" are used more or less interchangeably
when talking about neural networks. See, e.g., [this Stanford
course|http://cs231n.github.io/neural-networks-1/], "In other words, each
neuron performs a dot product with the input and its weights, adds the bias and
applies the non-linearity (or activation function)". Because the two terms are
interchangeable, I'm OK with either being used.
bq. If we think to apply this to more complex neural networks in the future, we
will need to support layers (ex. convolutional and pooling) other than the full
connected layer (ex. ReLU).
In my opinion, if this is a route Solr eventually wants to go, I think a better
strategy would be to just add a dependency on
[Deeplearning4j|https://deeplearning4j.org/]. New types of architectural
designs are constantly coming out (e.g.,
[attention|http://www.wildml.com/2016/01/attention-and-memory-in-deep-learning-and-nlp/],
which I believe [Deeplearning4j still does not even
support|https://deeplearning4j.org/roadmap]), and there are so many subtle ways
to vary neural networks that it would be a ton of work to try and maintain what
has effectively already been done in other libraries like Deeplearning4j.
> Implement RankNet.
> ------------------
>
> Key: SOLR-11597
> URL: https://issues.apache.org/jira/browse/SOLR-11597
> Project: Solr
> Issue Type: New Feature
> Security Level: Public(Default Security Level. Issues are Public)
> Components: contrib - LTR
> Reporter: Michael A. Alcorn
>
> Implement RankNet as described in [this
> tutorial|https://github.com/airalcorn2/Solr-LTR].
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
